What is the story of AI, where are we going?
It’s a question that was uncomfortable to me for a long time. Because, simply, I had no idea!
I probably still don’t, but I can give it a shot.
I stumbled upon AI-2027, an AI prediction by experts made in 2025. I loved reading it, because it told a well made, coherent story. But naturally AI has evolved differently.
Notably I’d say it’s going faster.
If you don’t want to get a bit spoiled, go read it first, https://ai-2027.com/.
We already have deceptive AI, not in intent, but in capability. And we are at the stage where AI has become strong enough to easily hack software with our current world (unless I underestimate the cyber budget, which I doubt) - I see it code without making mistakes besides security ones, ironically.
There is talk of the necessity of dialogue between China and the USA, like with the nuclear threat.
This may be cause for despair. But as a friend told me, we have the technology necessary to resolve the world’s issues if we focused them properly.
It’s easier than ever to write software
I just feel like capitalism is outdated.
Linkedin post on this:
Stop making AI better. Make it better.
When I was at the TRAIL gala when they asked: what do we really want to use AI for? It made me think.
Geertrui Mieke De Ketelaere said that experts should make the invisible visible. So let me try.
AI is a tool. Right now, it can automate workflows, connect to everything you own, deal with astronomical amounts of information, and take action on it. It can code, plan, critique, research, even hack, and all of that exceptionally well. Also, unlike code, anyone can query it in their own language.
That’s good enough for me. I mean this. Who needs it to be so good that it fully replaces us? I like being at the helm of my job, and I like my job. I must agree with Louis de Diesbach, I don’t want AGI.
We’re seeing AI described as “rogue” lately. But it’s not rogue. It just has new capabilities - like being better at deceiving, better at persuading - and it uses them. That’s not a bug, that’s what happens when you keep pushing for more powerful without pushing for more responsible. It’s powerful enough.
So what do I mean by “make it better”? Make it more energy efficient. More compressed. Better guardrails.
Take our time to learn the wise ways to use it (not only the optimal ways), and share them. Not rush to add it everywhere just because we can.
Here is one of the pre-prompts I give Claude: “take a deep breath, and do this step by step.” We could all use that advice.
You’re allowed to have an opinion on this. In fact, please do.
As for me, I don’t want to optimize the human out of the equation.
PauseAI is the organized version of this same stance — same worry, minus the wait-and-see.